On mercredi 6 décembre 2017 08:21:19 CET Michael Fritze wrote: > Remco, thanks for pointing on this result. > > Does anyone have a 1050 (Ti) and can confirm this result with current > Darktable version? Isn't the "boat" picture the same as the Bench.srw? > > 1060 can be in 3 and 6 GB versions. Is dt speed related to amount of card > memory? > 1060 is a bit higher priced than 1050 Ti, but when 2.5/3x faster it may be > worth it. passmark says approx 1.5x faster only. > 1050 Ti has several advantages: less power consumtion, smaller power suppy > possible, less noise that's why I considered this one. > Regards, Micha.
As Michael Rasmussen pointed out, the GPU isn't the only factor to take into account. You have to look at the whole system, to get the most for your money. I have the 1060 card (6 Gb version) on a system with an Intel i5-7500 and 16 Gb RAM. Not entry level, but certainly not top of the line either. The "boat" benchmark from that Phoronix page maxes out the GPU, not the CPU, and takes a few seconds to export, darkroom mode shows some delay on e.g. zooming in. But that image is particularly time-consuming, most of my own show no noticable delay while editing, even with the heavier operations like profiled denoise (or the heavier demosaic operations). And the fans on the card rarely kick in (certainly no worse than the system fans, probably because most of the time the GPU is used in short bursts). And I find that more important than the time taken on export: I usually edit a series of images, and then export those in a batch. While that runs, I can do something else, so no problem if it takes 10s or 20s. Having any noticable delay while actually editing is (for me) much more annoying, as it interrupts the flow (and even a fairly short delay gets annoying when you're adjusting a parameter). ____________________________________________________________________________ darktable user mailing list to unsubscribe send a mail to [email protected]
